Skip to content

Latest commit

 

History

History
21 lines (14 loc) · 1.03 KB

note-meeting-3.md

File metadata and controls

21 lines (14 loc) · 1.03 KB

What we have done

Divide our job into 4 parts

  • Built a sample for AES algorithm which can decrypt image in GPU with CUDA. (Jiemin)
  • Built a sample for Enclave that can send data to device memory. It will be used for storing the key for AES algorithm. (Mengxiao)
  • Implemented AES algorithm with python to encrypt data. (Chuyuan)
  • Built a sample for Pytorch which can make CUDA extension. (Xiaorui)

Tasks for next week

  • Combine all the parts and run it under Windows OS x64 with SGX.
  • It can use key stored in Enclave to decrypt encrypted image data Cifar in GPU.

What to expect

Next week, the enclave for storing keys and the GPU decryption algorithm will be shipped as a Pytorch extension. We will be able to test the system very soon.

Further ideas

  • From last meeting with instuctor, what if current GPU is not trustworthy when the whole system is compromised?
  • What can we do to make our idea still useful under assumption above? Rewrite a driver which can protect data in GPU?